The Forward Contract s Income Shifting Option. Mindy L. Mallory and Scott H. Irwin

Transcription

1 The Forward Contract s Income Shifting Option and Implications on the Forward Basis y Mindy L. Mallory and Scott H. Irwin Suggested cation format: Mallory, M. L., and Scott H. Irwin The Forward Contract s Income Shifting Option and Implications on the Forward Basis. Proceedings of the NCCC-134 Conference on Applied Commody Price Analysis, Forecasting, and Market Risk Management. St. Louis, MO. [

2 The Forward Contract s Income Shifting Option and Implications on the Forward Basis Mindy L. Mallory and Scott H. Irwin* Paper presented at the NCCC-134 Conference on Applied Commody Price Analysis, Forecasting, and Market Risk Management St. Louis, Missouri, April 19-20, 2010 Copyright 2010 y Mindy L. Mallory and Scott H. Irwin. All rights reserved. Readers may make veratim copies of this document for noncommercial purposes y any means, provided that this copyright notice appears on all such copies. *Mindy L. Mallory is an Assistant Professor and Scott H. Irwin is the Laurence J. Norton Chair of Agricultural Marketing, oth in the Department of Agricultural and Consumer Economics, Universy of Illinois at Urana-Champaign. The authors thank Hongxia Jiao and Gene Kunda for their assistance in collecting data for this study.

3 The Forward Contract s Income Shifting Option and Implications on the Forward Basis Previous studies have documented a cost of forward contracting grain relative to hedging in the futures markets. Our study quantifies the value of the income shifting option to forward contracting. An income shifting option refers to the fact that at harvest-time, a farmer can chose to sell uncontracted ushels of corn in the spot market or forward contract to sell after the first of the year. This option has non-trivial tax implications under a progressive tax system. By shifting income to the next tax year, a farmer can reduce the current year s income level and avoid a higher marginal income tax rate. Further, if country elevators have market power, they can capture some of the value of this income shifting option y offering a weak forward delivery January asis id. In a sufficiently captive draw area, an elevator knows that a farmer will e willing to accept a weak January forward asis id so long as he still receives some income tax enefs from deferring sales to the next tax year. This option is most valuale during years when farmer income is high. Therefore, in this study we posed that during years of high farmer income we would see forward asis ids which are anormally lower and appreciate at a slower rate than the harvest-time immediate delivery ids. We measure this effect using asis ids from elevators in seven regions in Illinois from 1980 to We find that a 1% increase in yield aove trend level decreases the January delivery forward asis ids y 3 cents per ushel; we also find that the January delivery forward asis ids appreciate at 44% the rate of the immediate harvest-time delivery asis ids. Keywords: Income tax, option value, marketing, asis, forward contract

4 The Forward Contract s Income Shifting Option and Implications on the Forward Basis Surveys show that farmers prefer forward contracting over futures contracts to manage price risk; e.g., see Musser, Patrick, and Eckman (1996) and Patrick, Musser, and Eckman (1998). Further studies estimate that there exists an implic cost of forward contracting; the cost of forward contracting which can e loosely defined as the change in the asis id from the time the contract is signed to the delivery date. See Brorsen, Coms, and Anderson (1995), Townsend and Brorsen (2000), and Shi et al. (2005) for examples. A farmer who has not previously contracted his grain at harvest-time has a different prolem, however. The aily to sell one s grain at harvest or enter into a forward contract wh January delivery provides the farmer wh an income shifting option. Suppose crop yield is relatively good in this particular year. Delivering all of one s grain efore January 1 st may result in the farmer paying income taxes from a higher racket; thus, having the option to transfer some or all of that income into the next calendar year is valuale. Presumaly an elevator makes a forward id ased on: forward id t = Futures price t T t Prof t where T t includes transportation and overhead costs and perhaps a risk premium. Elevators should e ale to id addional prof, Prof t, into the January forward id during high income years when the later delivery period allows farmers to transfer some income to the next calendar year. Previous research has given some attention to how tax policy effects farmer s marketing decisions. McNew and Gardner (1999) use a simulation model calirated to the U.S. corn market to examine how farmers storage ehavior changes under progressive and flat income tax systems. They find that carryover stocks are reduced and price variaily is increased under a progressive tax system relative to a flat tax system. Their insight is that under a progressive tax system, an increase in the inter tax-year price spread can induce less storage if the marginal taxrate is high enough. Tronstad (1991) explores after tax optimal hedging and storage ehavior through the cotton marketing year using a stochastic dynamic programming model. He finds that cash sales are preferred to storage early in the marketing year, ut as the end of the tax year approaches, storing cotton ecomes more attractive. This is ecause the enefs of deferring income to the next tax year outweigh the proaily of an adverse price movement. Tronstad and Taylor (1991) use a stochastic dynamic programming model to determine the optimal dynamic marketing strategy of a Montana winter wheat producer, where the producer can store, sell in the cash market, hedge in the futures market, or use a comination of these strategies. They find that when cash prices are low and efore tax income levels are low, cash grain sales are higher at the end of the tax year than at the eginning. Conversely, when efore tax income levels are high, cash sales are deferred until the next tax year and the price hedged in the futures market. 1

5 This ody of lerature is small, ut is consistent in s prediction that (progressive) income taxes influence a farmer s optimal storage ehavior. The question of whether or not this is reflected in actual farmer ehavior or in equilirium market outcomes has not een examined in actual data, however. In this article we recognize the aily to defer grain sales from one tax year to the next as a valuale income shifting option. It is highly likely that local grain elevator managers recognize the value of this option and would like to capture a portion for their own profs. The immediate delivery asis is calculated as the difference in the harvest-time corn spot price and the Decemer Chicago Board of Trade (CBOT) corn futures price, while the harvesttime forward asis id for January delivery is calculated as January delivery forward corn price (quoted at harvest) minus the March CBOT corn futures price. If elevator managers are ale to capture some of the income shifting option, we should find a specific pattern in spot and forward asis data. More specifically, farmer income should have a significant impact on the relationship etween the spot and forward ases. The first ojective of this study is to derive the theoretical value of the income shifting option ased upon the assumption that a farmer will prefer a January delivery forward contract to selling in the cash market at harvest so long as the forward spread is less in asolute value than the value of the income shifting option. The second ojective is to determine if the income shifting option is recognized (and capalized upon) y country elevator managers, and estimate how much this income shifting option weakens the January forward asis id relative to the harvest-time asis id. To this end we use a panel data set that contains forward and spot asis ids for harvest and January delivery at elevators in 7 regions in Illinois from the 1980 to 2009 crop years. The dataset was generated y a weekly survey of elevators throughout Illinois y the Illinois Ag Market News. In the next section we develop the conceptual framework of the income shifting option and show how can affect farmer decisions and equilirium asis levels. In the third section we descrie the data we employ. The following section descries our estimation strategy and a final section concludes. Conceptual Framework Denote Y as the realized yield for the current marketing year. Suppose that if the harvest-time price, P 0, is such that income in the current year is greater than some threshold, PY 0 > I, then the farmer is suject to a marginally higher income tax rate of τ. In high income years his net income is PY 0 ( 1 τ ). By delaying the sale of some of the crop a farmer can lower his income in the current year and avoid paying the extra tax. If we assume that he has not previously forward contracted any of his crop he will choose how much to sell in the cash market at harvest, Y 0, and how much to contract for January delivery, Y 1, ased on the cash and January delivery forward contract prices. Then a prof maximizing farmer will determine how much to sell in the spot market and how much to forward contract for January delivery according to the following decision prolem: 2

6 (1) max PY Y0, Y1 s.t. Y0 + Y1 = Y and ( 1 τ ) + PY if PY 0 0< I τ = > 0 if PY 0 0 I Consider the farmer s prolem if the harvest id and the January id were equal; i.e., P 0 = P 1. Then is clear that there is a strong incentive for the farmer to delay some of his income until January to avoid paying the addional income tax when τ > 0; specifically he will choose Y 0 so that P 0 Y 0 < I. Now suppose that the elevator manager recognizes this. Then the elevator can capture some of the income shifting enef y offering a January forward id that is discounted relative to the spot id y 0 < k < 1 so that P1 = kp0. Then the farmer s decision prolem ecomes: (2) max PY Y0, Y1 s.t. Y0 + Y1 = Y and ( 1 τ ) + kpy if PY 0 0< I τ = > 0 if PY 0 0 I Even when faced wh a discounted price the farmer will forward contract a posive amount for January delivery, selling only as much as Y0 = I P0at harvest, as long as the enef of doing so is greater than if he sells his entire crop in the spot market. In other words, as long as the elevator manager chooses k large enough so that (3) I P0 + kpy 0 1 PY 0 ( 1 τ ). P0 the farmer will accept the discounted forward price of P1 = kp0 and forward contract a posive amount for January delivery. Solving inequaly (3) for k, we see that as long as k is larger than the amount shown in inequaly (4) the farmer is willing to tolerate the price discount and still sell some of his crop in the next tax year: (4) PY 0 k ( 1 τ ) PY 0 1 I 3

7 Notice that this threshold is decreasing in the marginal income tax rate, τ. This means that as the marginal income tax rate increases, farmers are willing to accept a lower January forward id in exchange for the aily to shift some income into the next tax year. This concept implicly assumes that country elevators which uy corn from farmers have some level of market power. Otherwise competion among elevators would ensure the forward id was competively set relative to the spot asis and the enefs of the income shifting options would e enjoyed fully y the farmer. The lerature on market power of country elevators is sparse, ut Davis and Hill (1974) find some evidence of non-competive spatial price spreads among elevators in Illinois. A more recent case study of the Cargill and Continental merger y Hayenga and Wisner (2000) suggests that many farmers sell their grain whin a captive draw area. A captive draw area is an area around an elevator for which the transportation costs to an alternative location are high enough that the elevator is effectively a monopsony uyer of grain whin that area. In the next section we descrie the data set we use to determine if grain elevators can capalize on the income shifting option in the spot-forward asis differential. Data The panel dataset used in this study contains the asis for pre-harvest forward delivery contracts, harvest-time spot delivery, and harvest-time January forward contracts. The pre-harvest forward and harvest-time immediate delivery asis quotes are calculated relative to the CBOT Decemer futures contract. vest-time January forward asis quotes are calculated relative to the CBOT March futures contract. Bids are recorded on Thursdays in seven regions of Illinois over for corn and soyeans. The seven regions are Northern (1), Western (2), Northern Central (3), South Central (4), West Southwest (5), Waash (6) and Ltle Egypt (7). The asis ids are generated as a part of a daily survey y the Illinois Ag Market News of elevators throughout Illinois that conduct significant spot and forward transactions wh crop producers. The forward ases refer to #2 grade corn ought for shipment y rail or truck for harvest or January delivery to country elevators. Illinois Ag Market News currently disseminates the forward ases through daily electronic reports, ut, historical ases are pulished in a hard copy format only on a weekly asis. The range of ases in each of the seven regions is reported for forward ids on every Thursday. The mid-point of the reported high and low price is used to otain a single price for each region and week. The entire data set contains forward asis quotes for harvest delivery starting soon after the first of the year. For example, in the 1980 crop year, harvest forward asis ids are reported first on Feruary 22, 1980 and recorded weekly until the eginning of harvest, which in that year happens to e on Septemer 4, This is recognizale in the dataset since the pre-harvest forward ids cease and harvest-time immediate delivery ids egin starting on this date. Then near the end of harvest season Illinois Ag Market News starts to report the January forward asis ids. In a typical year this egins in mid-septemer to early Octoer and runs until the first part 4

8 of Decemer. The length of this period varies from year-to-year ecause depends on the eginning and duration of harvest. See tale 1 for a summary of definions and timeframes of the data in our sample. For our analysis, is the rief period during harvest when immediate delivery asis ids and January delivery forward asis ids are simultaneously offered that will help us answer our question of interest. Therefore, we construct our dataset using only dates when oth an immediate delivery asis id and a January delivery forward asis id are quoted. Depending on the length of harvest this means that some crop years have more oservations than others. In the next section, we explore the properties of the data set prior to specifying an econometric model to test the income shifting hypothesis. Evidence of the Income Shifting Option in the Data In figure 1 we present a sample of the data. This shows a scatter-plot of the forward and spot asis ids for region 4 which is South Central Illinois. Three different asis ids are pooled across time and displayed on this chart. Pre-harvest forward asis ids are represented y the lue triangles and lue trend line. The red squares denote the harvest-time immediate delivery asis ids; the trend line is in red. January forward asis ids are represented y the green triangles and green trend line. In this chart the two time periods, pre-harvest and harvest are visile in the patterns of the data. It is instructive to focus on the trend lines for a moment, however. Figure 2 contains the same information as figure 1, ut wh the individual data points suppressed so that we can focus on the trends. The pre-harvest (lue line) period has een the focus of the cost of forward contracting lerature. The upward slope is indicative of the cost of forward contracting. By cost of forward contracting we mean that if a farmer signs a forward contract for harvest delivery he will, in an average year, receive a weaker asis that if he had hedged in the futures market and sold his actual grain in the cash market at harvest-time. The prolem we focus on in this paper, however, pertains to the harvest-time ehavior ecause this is the period that can shed light on the income shifting option s effect. The green line in figure 2 represents the time trend of the forward asis for January delivery as quoted during the harvest season. The red line is the time trend of the immediate delivery harvest-time asis ids. Compare the trend line of the forward asis for January delivery to the trend line of the immediate delivery harvest-time asis. The immediate delivery asis id is noticealy stronger on average than the forward asis id. Since the January forward contract is deliverale just days later and since all uncertainty aout the size of the harvest, typically has een resolved at this point, we suggest that the relative asis pattern in the days just efore the first of the year is consistent wh our theory that elevator managers can capture at least some of the income shifting option. 5

9 The slopes of the two trend lines also are informative. The harvest-time immediate delivery quotes have a steep time trend relative to the January delivery forward asis quotes. The steep upward slope of the harvest delivery ids is not surprising given the glut in the local spot market when harvest egins. The relative slopes of the harvest-time immediate delivery trend and the January delivery forward asis trend is suggestive of the value of the income shifting option. The immediate delivery harvest-time ids see an improvement of approximately $0.30 per ushel during the harvest period, while the January forward asis ids strengthen y only aout $0.15 per ushel during the same time period. This dwarfs estimates of the pre-harvest cost of forward contracting which are generally on the order of magnude of $0.05 per ushel in the wheat, corn and soyean markets. The discussion aove has limed scope, however, ecause figures 1 and 2 display the data pooled over time. In any given crop year these patterns can e significantly different, and the average values may not tell the whole story. After all, the harvest-time asis pattern varies according to yield levels, carryover stocks, and any numer of local supply and demand condions. Figures 3-9 show the pre-harvest and harvest asis ids for regions 3 and 4 in five year intervals starting wh 1980 through Regions 3 and 4 cover much of central Illinois, a region located in the heart of the Corn Belt. This shows that the asis does not always improve during the harvest season. Notice however, that when the time trends are posive during the harvest period, the slope of the immediate delivery harvest-time trend is always larger than the slope of the January forward delivery trend. Also, when the slopes are negative, the January forward delivery ids are stronger than the immediate delivery ids. This seems to e evidence that the degree to which elevator managers are ale to capalize on the value of the income shifting options is sensive to the condions of the local market. Un Root Testing Before we specify and test an econometric model of the income shifting option we need to characterize the asis data. The spot (forward) asis data are constructed as the difference etween the spot (forward id) and futures price at time t. If the futures market is functioning well, the spot and futures prices and the forward ids and futures prices should e highly integrated. This means that movement in the futures prices should imply movement in the spot prices and forward ids. This loosely implies that we can test for a un root in the asis data to determine if the markets were functioning well enough to conduct this kind of analysis. Stationary in the asis indicates the spot and futures markets are well integrated. We also need to determine the stationary of the asis variales to conduct the econometric analysis elow. We suject the panel of asis data to a typical attery of un root tests, the results of which are reported in tale 1. We use StataIC 11 to conduct the Levin-Lin-Chu, ris-tzavalis, Breung, and Im-Pesaran-Shin tests as well as the Fisher-type tests from Choi (ris and Tzavalis 1999; Breung 2000; Choi 2001; Levin, Lin and Chu 2002; Im, Pesaran and Shin 2003; Breung and Das 2005). These tests all investigate the null hypothesis that the panels contain un roots against an alternative hypothesis that the panels are stationary. Where appropriate, lag 6

10 lengths are selected y the AIC. Each un root test we conduct rejects the un root hypothesis, so we proceed assuming the asis variales are stationary. Estimation The conceptual framework along wh the casual examination of the asis data motivates a specification for an econometric model relating the harvest-time immediate delivery asis to the January delivery forward asis. Both suggest that the harvest-time immediate delivery asis is the primary determinant of the January forward asis id. The January delivery forward asis is distinguished, however, y condions in the spot market. In the conceptual framework we pos that in years when there is a large crop farmer income will e higher than average which increases the value of the income shifting option. This should have a depressing effect on the January forward asis. Using Percent Deviation from Trend Yield to Measure the Value of the Income Shifting Option Motivated y the previous discussion we specify an econometric model relating the harvest-time immediate delivery asis to the January delivery forward asis y equation (5). Jan (5) = β0 + β1 + β2yt + ui + ε where Jan is the January delivery asis id in region i at time t, immediate delivery spot asis id in region i, u i is a region level effect, t is the harvest-time y is the percent deviation from detrended mean yield for that crop year, and ε is the usual random error term. The data we examine in this specification only consists of time periods t where there existed a spot harvest-time id and a January forward asis id on the same day. Forward asis ids for January delivery rarely are offered early enough to overlap wh pre-harvest forward asis ids, so we restrict our analysis to spot harvest-time ids and January forward asis ids. The econometric model assumes the January forward asis id in region i is determined y three factors. The current cash asis is the primary determinant. The constant term captures the average level of compensation for the risk of holding the contract until January. This is different from the market risk of an adverse price movement ecause if a risk premium for price variaily existed would already e emedded in the spread etween the spot price and the March futures price, which is emedded in the two ases. The same is true of any market return to storage. Therefore β 0, if is significantly different from zero, reflects risk each counterparty entertains y holding a contract that matures some weeks in the future. The final component of the econometric model is the realized crop yield. We postulate that the elevators are ale to use this measure, which is known wh reasonale accuracy even as harvest is still ongoing, as a measure of farmer s income associated wh this year s harvest. In this way the elevator can predict years when the value of the income shifting option is high and consequently offer a weaker January asis id. 7

11 This motivates two hypothesis tests. The first is a null hypothesis that β 2 = 0, which would imply that the January forward id is not influenced y farmer income level in a particular year. The second is a null hypothesis that β 1 = 1; this would imply that the January forward asis fluctuates perfectly wh the harvest-time immediate delivery asis. While β 1 < 1 indicates that the January forward asis does not appreciate as much on average as does the harvest-time immediate delivery asis during the harvest season. We find the region specific fixed effects, u i, are necessary in the model ecause the Breusch and Pagan (1980) test rejects the null hypothesis of no individual effects wh a test 2 statistic of χ = and a p-value of p = We choose a fixed effects rather than a random effects model ecause the Hausman (1978) specification test rejects the null hypothesis that the individual effects are uncorrelated wh the other regressors wh a test statistic of 2 χ = and a p-value of p = This model undoutedly contains a high degree of autocorrelation. We could account for the autocorrelation y estimating the model as a dynamic panel, which involves instrumenting wh lags of the dependent variale. However, the dynamic panel estimation procedure of Arellano (1990) and extended in Arellano and Bond (1991) was designed specifically for data sets that have a large numer of panels and short numer of time periods (Greene 2003). Our suation is exactly the oppose, small n and large T, wh n = 7 and T =343, since we only have 7 regions in Illinois and weekly data from the 1980 to 2009 crop years. Instead we include a lag of the harvest cash asis id to account for the autocorrelation and use the fixed effects (LSDV) estimator. We estimate all models in StataIC 11 using the xtreg fixed effects routine. Regression results are found in tale 3. The top panel of tale 3 contains the econometric model as specified in equation (5) ut also including the lagged value of the harvest-time forward asis id, 1, to account for the autocorrelation. We had 2,371 oservations and 7 regions whin Illinois, each region contained roughly 335 oservations. All variales in this model are significant at the 5% level and all ut the lagged harvest id are significant at the 1% level. The sign on the corn yield deviation is negative, which is consistent wh our first hypothesis that in years of ounty elevators are ale to offer a lower January forward id than they otherwise would, capalizing on the farmer s income shifting option in forward contracting for January delivery. Since the un of the corn yield variale is in percent deviation from the detrended mean, we can interpret the marginal effects in the natural way. If corn yields are 1% higher than trend, the model predicts the January forward asis id will e 3 cents less than if yields were at trend levels, all else equal. On 5,000 ushels of corn, this amounts to $150 for every 1% that yield is aove trend level. The second hypothesis concerns the estimate of β 1, which is 0.44 in this model and significantly different from 1. This reflects the scatter plot of the data in figures 1 and 2 where the slope of the January delivery forward asis trend line appears to e approximately half that of the harvest-time immediate delivery trend line. A coefficient estimate of 0.44 implies that the January forward asis will move in the same direction as the harvest-time immediate delivery 8

12 asis, uy only 44% as much. This means that if the immediate delivery asis id experiences a $0.30 per ushel improvement over the course of the harvest season, the January delivery forward asis id will only see a 13 cent per ushel improvement. On 5,000 ushels of corn this amounts to a $660 cost of forward contracting relative to selling in the cash market. In the ottom panel of tale 3 we report regression results for the same analysis as reported in the top panel ut whout 1 as a roustness check. The results are similar wh the coefficient on yt negative and significant and the coefficient on i significantly different than 1. Using the Value of the vest to Measure of the Value of the Income Shifting Option Using yield alone as a proxy for farmer income is a concern since the natural hedge ased on the negative movement of equilirium price and yield is not negligile. To account for this we alter how we measure the value of the realized harvest. We define the variale I = s Yt. This functions as a proxy for farmer income; the spot price in region i at time t is s, and the detrended yield level for crop year t is Y t. The new model is contained in equation (6). = β + β + β I + v + ε Jan (6) i Again the Breusch and Pagan test rejects the null hypothesis of no individual effects and the Hausman speciation test rejects the null hypothesis that the individual effects are uncorrelated 2 wh other regressors wh a test statistic of χ = and p = 0.000, so we estimate this model as well as a fixed effects model. Tale 4 contains the regression results from this specification. As wh the previous specification, the top panel displays the estimation results from a model which contains the lagged harvest asis, 1, while the ottom panel does not. The coefficient on farmer income is negative and significant at the 1% level, which is consistent wh the previous model and our hypothesis. The effect of a 1% increase in farmer income translates into a 2 cent per ushel weakening of the January asis id, all else equal. The estimate of the coefficient on β 1 is 0.45 and is significantly different from 1. Limations The analysis aove comes wh some limations. First, we are using asis ids, which do not necessarily mean that any transactions took place at these prices. This creates prolems ecause is not clear how well zero-transaction ids reflect the true market price at a local elevator. Further, the asis ids in our data set are only for forward contracts. Bids on other types of contracts do not appear in our data set. For example, delayed pricing contracts do not appear in 9

13 our sample; therefore, they could e an important component of farmers income tax strategy that we do not oserve. A second limation to our analysis is that we estimate a reduced form model which has no ehavioral variales in the specification. This means that we can only demonstrate that the data are consistent wh our income shifting theory, ut we cannot assert that these price relationships are in fact driven y the factors we propose. There are other plausile factors which could contriute to this kind of spot-forward asis relationship. For example, the end of the tax year coincides wh a major holiday season, falling etween Christmas and New Year s Day. It is possile that spot-forward asis patterns are influenced y a seasonal holiday slowdown of grain delivery and contracting wh a susequent increase after the first of the year. However, since we did find that farmer income was a significant factor in determining the spotforward asis relationship we dout the holiday is what is driving the patterns oserved in the data. A grain elevator has s own risk management and income tax considerations which could affect the way spot and forward asis ids are set. It is possile that these considerations drive the results we find in the econometric model more than our income shifting hypothesis. Conclusions Previous studies have documented a cost of forward contracting grain relative to hedging in the futures markets. Our study quantifies the value of the income shifting option to forward contracting. An income shifting option refers to the fact that at harvest-time, a farmer can chose to sell uncontracted ushels of corn in the spot market or he can forward contract to sell them after the first of the year. This option has non-trivial tax implications under a progressive tax system. By shifting income to the next tax year, a farmer can reduce the current year s income level and avoid a higher marginal income tax rate. Further, if country elevators have market power, they can capture some of the value of this income shifting option y offering a weak forward delivery January asis id. In a sufficiently captive draw area, an elevator knows that a farmer will e willing to accept a weak January forward asis id so long as he still receives some income tax enefs from deferring sales to the next tax year. This option is most valuale during years when farmer income is high. Therefore, in this study we posed that during years of high farmer income we would see forward asis ids which are anormally lower and appreciate at a slower rate than the harvest-time immediate delivery ids. We measure this effect using Illinois Ag Market News asis ids from elevators in seven regions in Illinois from 1980 to We find that a 1% increase in yield aove trend level decreases the January deliver forward asis ids y 3 cents per ushel; we also find that the January delivery forward asis ids appreciate at 44% the rate the harvest-time immediate delivery asis ids appreciate. These findings are consistent wh our hypothesis that country elevators are ale to capture some of the income shifting enefs of forward contracting through the relative immediate delivery harvest and January delivery forward ases. 10

14 We would like to confirm the income shifting hypothesis further. Our data set currently does not contain March delivery forward ases. However, if we could compare the pattern of immediate delivery harvest and January delivery forward ases wh the pattern of immediate delivery January and March delivery forward ases we could learn a lot. It makes no difference on a farmer s income tax ill if he delivers in January or March, ut does when he decides etween Decemer and January delivery. 11

15 References Arellano, M. (1990). "Testing for Autocorrelation in Dynamic Random Effects Models." Review of Economic Studies: Arellano, M., and S. Bond (1991). "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations." Review of Economic Studies 58: Breung, J. (2000). "The Local Power of Some Un Root Tests for Panel Data." Advances in econometrics 15: Breung, J., and S. Das (2005). "Panel Un Root Tests under Cross-Sectional Dependence." Statistica Neerlandica 59: Breusch, T., and A. Pagan (1980). "The Lagrange Multiplier Test and Its Applications to Model Specification in Econometrics." Review of Economic Studies 47: Brorsen, B., J. Cooms, and K. Anderson (1995). "The Cost of Forward Contracting Wheat." Agriusiness 11: Choi, I. (2001). "Un Root Tests for Panel Data." Journal of International Money and Finance 20: Davis, L., and L. Hill (1974). "Spatial Price Differentials for Corn among Illinois Country Elevators." American Journal of Agricultural Economics 56: Greene, W. (2003). Econometric Analysis, prentice Hall Upper Saddle River, NJ. ris, R., and E. Tzavalis (1999). "Inference for Un Roots in Dynamic Panels Where the Time Dimension Is Fixed." Journal of Econometrics 91: Hausman, J. (1978). "Specification Tests in Econometrics." Econometrica 46: Hayenga, M., and R. Wisner (2000). "Cargill's Acquision of Continental Grain's Grain Merchandising Business." Review of Agricultural Economics 22: 252. Im, K., M. Pesaran, and Y. Shin (2003). "Testing for Un Roots in Heterogeneous Panels." Journal of Econometrics 115: Levin, A., C. Lin, and C. Chu (2002). "Un Root Tests in Panel Data: Asymptotic and Fine- Sample Properties." Journal of Econometrics 108: McNew, K., and B. Gardner (1999). "Income Taxes and Price Variaily in Storale Commody Markets." American Journal of Agricultural Economics 81:

16 Musser, W., G. Patrick, and D. Eckman (1996). "Risk and Grain Marketing Behavior of Large- Scale Farmers." Review of Agricultural Economics 18: Patrick, G., W. Musser, and D. Eckman (1998). "Forward Marketing Practices and Attudes of Large-Scale Midwestern Grain Producers." Review of Agricultural Economics 20: Shi, W., S. Irwin, D. Good, and S. Dietz (2005). "Wheat Forward Contract Pricing: Evidence on Forecast Power and Risk Premia." Proceedings of the NCCC-134 Conference on Applied Commody Price Analysis, Forecasting, and Market Risk Management, St. Louis, MO. Townsend, J., and B. Brorsen (2000). "Cost of Forward Contracting d Red Winter Wheat." Journal of Agricultural and Applied Economics 32: Tronstad, R. (1991). "The Effects of Firm Size and Production Cost Levels on Dynamically Optimal after-tax Cotton Storage and Hedging Decisions." Southern Journal of Agricultural Economics 23. Tronstad, R., and C. Taylor (1991). "Dynamically Optimal after-tax Grain Storage, Cash Grain Sale, and Hedging Strategies." American Journal of Agricultural Economics 73:

17 Tale 1: Data Definions Types of Basis Bids Time Period Bids are Offered Futures Contract Bases are Calculated Against Pre-vest Forward Jan/Fe-Beginning of vest Decemer vest-time Immediate Delivery Beginning of vest-end of vest Decemer January Forward Beginning of vest-end of vest March 14

18 Tale 2: Panel Un Root Tests For Basis Variales a Levin-Lin-Chu adjusted t p-value Lags Jan ris-tzavalis ρ z p-value Jan Breung λ p-value Jan Im-Pesaran-Shin Z tar p-value Lags Jan Fisher-type Inverse χ 2 P p-value Lags Jan Inverse Normal z p-value Lags Jan Inverse Log L* p-value Lags Jan a All tests examine a null hypothesis that the panels contain un roots, against the alternative of stationary panels. Where applicale, lags are chosen y the AIC. 15

19 Tale 3: Corn yield and the income shifting option in the January forward asis R 2 whin = Os = 2371 etween = Groups = 7 overall = Os per group F(3, 2361) = min = 332 Pro > F = avg = max = 340 Coef. Std. Error t P > t I cons σ u σ e ρ fraction of variance due to u i R 2 whin = Os = 2372 etween = Groups = 7 etween = Os per group F(2, 2363) = min = 333 Pro > F = avg = max = 340 Coef. Std. Error t P > t I cons σ u σ e ρ fraction of variance due to u i 16

20 Tale 4: Corn per acre income and the income shifting option in the January forward asis R 2 whin = 0.68 Os = 2371 etween = 0.92 Groups = 7 overall = 0.64 Os per group F(3, 2361) = min = 332 Pro > F = avg = max = 340 Coef. Std. Error t P > t I cons σ u σ e ρ fraction of variance due to u i R 2 whin = 0.67 Os = 2372 etween = 0.93 Groups = 7 overall = 0.66 Os per group F(2, 2363) = min = 333 Pro > F = avg = max = 340 Coef. Std. Error t P > t I cons σ u σ e ρ fraction of variance due to u i 17

21 Figure 1: Preharvest and vest Forward Bases in Central Illinois, Pooled Across Years 18

22 Figure 2: Trendlines of Preharvest and vest Forward Bases in Central Illinois, Pooled Across Years 19

23 Figure 3: Preharvest, vest, and Post vest Forward Bases in Central Illinois,

24 Figure 4: Preharvest, vest, and Post vest Forward Bases in Central Illinois,

25 Figure 5: Preharvest, vest, and Post vest Forward Bases in Central Illinois,

26 Figure 6: Preharvest, vest, and Post vest Forward Bases in Central Illinois,

27 Figure 7: Preharvest, vest, and Post vest Forward Bases in Central Illinois,

28 Figure 8: Preharvest, vest, and Post vest Forward Bases in Central Illinois,

29 Figure 9: Preharvest, vest, and Post vest Forward Bases in Central Illinois,